README.md

---
tags:
- creativityneuro
- llm-creativity
- mechanistic-interpretability
base_model: microsoft/Phi-3.5-mini-instruct
license: apache-2.0
---

# phi-3.5-mini-instruct-cn-problem-kr0.2-a0.01-creative

This is a **CreativityNeuro (CN)** modified version of [microsoft/Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct).

## Model Details

- **Base Model**: microsoft/Phi-3.5-mini-instruct
- **Modification**: CreativityNeuro weight scaling
- **Prompt Set**: problem
- **Keep Ratio**: 0.2 (top 20.0% of task-specific weights)
- **Alpha**: 0.01 (scaling strength)
- **Mode**: creative

## What is CreativityNeuro?

CreativityNeuro identifies task-specific neurons using Wanda-style importance scoring and selectively
upscales weights associated with creative thinking. The modification formula is:

```
W_new = W × (1 + α × mask)
```

Where `mask` identifies weights important for creative tasks but not for routine/associative tasks.

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("priorcomputers/phi-3.5-mini-instruct-cn-problem-kr0.2-a0.01-creative")
tokenizer = AutoTokenizer.from_pretrained("priorcomputers/phi-3.5-mini-instruct-cn-problem-kr0.2-a0.01-creative")

# Use like any other model
outputs = model.generate(...)
```

## Citation

If you use this model, please cite:

```bibtex
@misc{creativityneuro2025,
  title={CreativityNeuro: Mechanistic Interpretability for LLM Creativity},
  author={Prior Computers},
  year={2025},
  url={https://huggingface.co/priorcomputers}
}
```